It appears that the IPUMS group has created the HHTYPE = 9 (HHTYPE could not be determined). As I understand it, this group is mostly for unmarried partners. We were reviewing this category and poverty and came upon a household that does not seem to make much sense in regard to the family relationships and poverty. I am wondering if this has something to do with the recoding for HHTYPE=9 or if there is something else going on.
I am including a table to show the data. But I will give a short description. The household records include 5 people- 2 adults and 3 children. It is a HHTYPE = 9. The head of household is a female (rec 1). There is one family and one subfamily. The head of household is not part of the subfamily. There is a male adult (rec 5) who is identified as an unmarried partner who is in the subfamily - along with all 3 children. The children’s mother is identified as record 1 and father as record 5. The head of household has no income. The unmarried partner has income. The head of household and children have poverty level =1. The unmarried partner has poverty level 245.
There are two related questions.
1 Why is subfamily described as married couple with children when it is the father (Unmarried partner and 3 children).
2 Why are the children given the mother’s income when they are defined as the father’s sub family.
You are correct that HHTYPE is implicated in the response, but it is also related to how IPUMS assigns spouse locations and the different ways of measuring families and subfamilies. I will respond to each of your specific questions, but also share a bit of summary information.
The subfamily is described as a “married couple” with children in SFTYPE because the unmarried partner has a spouse/partner assigned via SPLOC. The logic that defines sub family types looks at SPLOC rather than marital status.
FTOTINC is based on FAMUNIT, not SUBFAM (which is what I assume you are showing in the sample household). All members of this household have the same FAMUNIT value; in contrast, SUBFAM identifies the householder as one subfamily and the unmarried partner and children as another subfamily. For persons in families where HHTYPE is greater than 4, all persons in the same FAMUNIT are assigned the INCTOT value of the first person. In the case you mention, this is the female householder with an income of 0. The documentation doesn’t specify how FTOTINC relates to HHTYPE; we will clarify this relationship in the documentation in a future release.
What is less clear to me is why/how all records are not defined as the same sub-family (and how the children of the householder are associated with the unmarried partner in the subfamily rather than the parent who is explicitly identified by their RELATE value); I have discussed this briefly with the IPUMS USA team. They have plans to do a systematic revamping of these variables in the next year or two.
In the meantime, you use RELATE and the family pointer variables available from IPUMS: MOMLOC, POPLOC, SPLOC to implement family/subfamily classifications in a way that seems appropriate for your research.
This is a function of how the Census Bureau defines families for the official poverty measure (OPM). Unmarried partners are not considered part of the resource sharing unit; as such, their income to poverty ratio is calculated separately from others who may appear to be in their family based on values for RELATE, FAMUNIT, SUBFAM, or CBSUBFAM. You can see this resource on poverty measures from the University of Wisconsin and this Census Bureau report on income and poverty for more information. I have listed the household showing the original value for poverty to income ratio from the ACS PUMS data (us2018c_povpip)—you can see that the father/unmarried partner has a separate value in the original data as well as the IPUMS variable POVERTY, while the mother and her children have identical values as they are a family/resource unit based on OPM definitions. Clearer documentation of this would be helpful. Until we have completed a thorough review and revamp of the family measures and updated the documentation, I recommend creating these family-level measures based on the family/subfamily classifications that are appropriate for your research with an eye to different census definitions of poverty described in the resources linked above.
